Data Centre Power Use Could Hit 945 TWh by 2030 as AI Waste Spurs Calls for Restraint
Updated
Updated · The Conversation · Jun 17
Data Centre Power Use Could Hit 945 TWh by 2030 as AI Waste Spurs Calls for Restraint
3 articles · Updated · The Conversation · Jun 17
Summary
945 terawatt-hours of annual electricity use by data centres could be reached by 2030, the report says, as routine AI queries add to carbon emissions, water demand, grid strain and e-waste.
Simple tasks such as drafting short messages or polishing sentences often do not need large AI models; clearer prompts, shorter outputs and less use of image, audio and video tools can cut unnecessary computing.
Large organisations are urged to avoid adding AI to products by default and to test whether smaller models or non-AI tools can solve the same problem.
Governments are also pressed to require data centres to disclose electricity use, water use, emissions and e-waste, with Australia cited as relying on voluntary guidelines rather than binding rules.
The broader argument is that AI should be treated like any other utility—useful, but carrying visible environmental costs that need to become part of normal governance and daily habits.
Can AI solve climate change before its own energy thirst worsens it?
Is your daily AI habit secretly fueling a global environmental crisis?
Powering AI: The Escalating Resource and Community Impacts of Data Centers to 2030
Overview
The rapid advancement and widespread adoption of artificial intelligence are dramatically increasing global resource demands, especially in data centers, which are the backbone of AI operations. As these centers become more resource-intensive, they drive significant consumption of electricity, water, and raw materials. Much of the electricity needed is still sourced from fossil fuels, and projections show that AI is a dominant force behind the rising energy needs of data centers, contributing to a broader increase in global electricity demand. This trend highlights the urgent need for sustainable solutions as we approach 2030.